Multiple Modes of Repression by the Smad Transcriptional Corepressor TGIF*

TGIF is a DNA-binding homeodomain protein that has been demonstrated to play a role in transforming growth factor β-regulated transcription and implicated in the control of retinoid-responsive transcription. We investigated the intrinsic transcriptional activity of TGIF fused to a heterologous DNA-binding domain. Our results demonstrate that TGIF is a transcriptional repressor able to repress transcription from several different promoters. Repression by TGIF is insensitive to the distance at which it is bound from the promoter. Moreover, the wild type TGIF effectively represses transcription when bound to its cognate DNA-binding site via its homeodomain. Deletion analysis revealed the presence of at least two separable repression domains within TGIF. Repression by one of these is dependent on the activity of histone deacetylases (HDACs), whereas the other appears not to require HDAC activity. Finally, we demonstrate that TGIF interacts with HDACs via its carboxyl-terminal repression domain. Together, these results suggest that TGIF is a multifunctional transcriptional repressor, which acts in part by recruiting HDAC activity.

TGIF is a DNA-binding homeodomain protein that has been demonstrated to play a role in transforming growth factor ␤-regulated transcription and implicated in the control of retinoid-responsive transcription. We investigated the intrinsic transcriptional activity of TGIF fused to a heterologous DNA-binding domain. Our results demonstrate that TGIF is a transcriptional repressor able to repress transcription from several different promoters. Repression by TGIF is insensitive to the distance at which it is bound from the promoter. Moreover, the wild type TGIF effectively represses transcription when bound to its cognate DNA-binding site via its homeodomain. Deletion analysis revealed the presence of at least two separable repression domains within TGIF. Repression by one of these is dependent on the activity of histone deacetylases (HDACs), whereas the other appears not to require HDAC activity. Finally, we demonstrate that TGIF interacts with HDACs via its carboxyl-terminal repression domain. Together, these results suggest that TGIF is a multifunctional transcriptional repressor, which acts in part by recruiting HDAC activity.
Homeodomain transcription factors play roles in both transcriptional repression and activation, and their action is critical for the correct regulation of diverse developmental processes (1). Homeobox genes, first discovered in Drosophila (2,3), where they regulate segment identity, have now been identified in yeast, plants, and many metazoans. The homeodomain is a conserved 60-amino acid DNA-binding domain consisting of three ␣-helices, the third of which is the major DNA recognition helix (4). In contrast, helices one and two appear to be less important for DNA binding but play a role in specific interactions with other proteins. Thus, the homeodomain proteins HoxB1 and Pbx1 have been shown to bind to DNA together, and they interact with each other primarily via helices one and two (5). In addition, the interaction of homeodomain proteins with components of the basal transcription machinery has also been shown to be mediated via the amino-terminal part of the homeodomain (6,7).
Homeodomain proteins can be grouped into multiple subfam-ilies, based on such criteria as the primary sequence of the homeodomain, their flanking sequences, and organization into gene clusters (1). One subfamily of homeobox genes encodes proteins with an atypical homeodomain referred to as TALE (for three-amino acid loop extension) homeodomains (8,9). This class of proteins has a three-amino acid loop between helices one and two of the homeodomain. This insertion does not appear to affect DNA binding but may play a role in determining specific interactions with other transcription factors (5). Members of this subfamily have been identified in diverse species (9) and include the yeast Mat␣2 protein (10) and the human proteins Pbx1, Meis1, and TGIF (8,(11)(12)(13). Human TGIF was cloned by its ability to bind to a specific retinoid-responsive element, and its DNA binding specificity has been determined (8). TGIF mRNA has been shown to be present in a wide variety of adult human tissues, although expression levels vary greatly. Expression of TGIF appears to be most strictly regulated within the brain (8). Murine TGIF transcripts are absent from the early embryo and are first detectable at embryonic day 9.5. In situ analysis of TGIF expression in mouse embryos revealed expression in an extensive range of tissues, including the brain. However, during development, expression in the brain becomes restricted to specific areas of proliferating cells, such as the cerebellar neuroepithelium (14). Human TGIF maps to 18p11.3 (15), a region that contains a locus (HPE4) associated with holoprosencephaly (16), a prevalent congenital disorder of brain and craniofacial malformation (17).
Transcriptional repression can be brought about in various ways. One way is by competition for binding to DNA, thus displacing activators (18). In the case of TGIF, its binding to the retinoid X receptor response element within a reporter construct inhibits retinoid X receptor-dependent transcription in this manner (8). Transcriptional regulation can also be achieved by the remodeling of chromatin structure by histone acetyl transferases and histone deacetylases (HDACs) 1 (19,20). The recruitment of HDACs by transcriptional repressors to specific target genes results in the compaction of nucleosome structure, limiting accessibility of the DNA template to transcription factors (21)(22)(23)(24)(25)(26)(27). TGIF and the mediator of TGF-␤ signaling, Smad2, have recently been shown to interact following TGF-␤ receptor activation (28). This interaction results in the recruitment of TGIF to TGF-␤ response elements, repressing TGF-␤-activated transcription. Repression of Smaddependent transcription by TGIF correlates with the recruitment of HDAC instead of the coactivator p300 into the Smad complex (28 -31). Together, these results suggest that TGIF acts to repress transcription. However, little is known about whether TGIF has intrinsic transcriptional repressor activity or whether TGIF is a context-dependent repressor. Here we show that TGIF is an active transcriptional repressor. This transcriptional repression appears to be dependent on multiple regions of the TGIF protein and, at least in part, requires interaction of TGIF with a histone deacetylase.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-Mink lung epithelial L17 cells were cultured in minimum Eagle's medium supplemented with 10% fetal bovine serum, and COS-1 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. L17 cells were transfected in 6-well plates using DEAE-dextran, as described previously (32).
Reporter Assays-Firefly luciferase was measured using Promega luciferase assay reagents, and Renilla luciferase was assayed in 25 mM Tris (pH 7.5) and 100 mM NaCl with 0.09 M coelenterazine (Biosynth). pCMV-Renilla luciferase (Promega) was included in all transfections to control for transfection efficiency. Five Gal4p binding sites are present upstream of the E1b TATA box in (Gal) 5 -E1b-luc. The (Gal) 5 -AdMLPluc, (Gal) 5 -SV40-luc, and its derivatives with intervening sequence between the Gal4p sites and the SV40 promoter are as described previously (33). (Gal) 4 -TK-luc has four Gal4p binding sites upstream of the TK promoter from pBLCAT2 (34), driving luciferase. (TG) 5 -TK-luc has five copies of a TGIF binding site (gatctcCTGTCAAtaca (TGIF binding site in uppercase letters)) inserted into the BamHI site of TK-luc. L17 cells were transfected with the indicated reporters (1000 ng/6-well plate) together with the indicated expression constructs. Total amounts of DNA were kept constant with either pCMV5 lacking an insert or a pCMV5-farnesyl transferase construct. TSA (in 100% ethanol) was added to the indicated concentrations 18 h before cell lysis.
Expression Constructs-TGIF expression constructs were created within a modified pCMV5 containing a single Flag epitope or two copies of a HA epitope. Deletion constructs were created by polymerase chain reaction or by using existing restriction sites within the TGIF sequence. HDAC expression vectors are as described previously (35). GBD fusions were created within a Gal4p DNA-binding domain vector (pM; CLONTECH).

TGIF Represses
Transcription-To assess the role of TGIF in transcriptional regulation, we fused its entire coding sequence to the GBD. This expression construct was transfected into mink lung epithelial (L17) cells, together with various luciferase reporter constructs. When coexpressed with a reporter in which transcription of the luciferase gene was driven by a minimal TATA element and multiple Gal4p binding sites, GBD/TGIF failed to activate transcription. In contrast, GBD/ TGIF repressed transcription from this reporter when compared with the activity obtained in the presence of a control plasmid or GBD alone (Fig. 1A). To determine whether TGIF was able to repress transcription from more complex promoters, GBD/TGIF was targeted via multiple Gal4p sites to reporter constructs in which transcription was activated by the herpes simplex virus TK promoter or the adenovirus major late promoter. In addition to a TATA element, binding sites for Sp1 and NF-1 are present within the TK promoter, and the adenovirus major late promoter contains a CCAAT box and a binding site for USF (34,36). The activities of both the adenovirus major late and TK promoters were dramatically repressed by targeting of TGIF (Fig. 1, B and C). In contrast, little effect of GBD alone was observed.
Repression from a Distance-To determine whether this repression was dependent on the distance of the binding sites from the promoter region, TGIF was targeted to multiple Gal4p sites at different distances from the SV40 promoter. As shown in Fig. 1D, luciferase expression driven by the SV40 promoter was strongly repressed by GBD/TGIF. In addition, when the Gal4p sites were separated from the promoter region by either 450 base pairs or 2.9 kilobase pairs, repression by GBD/TGIF was essentially unchanged. Thus, TGIF appeared to be able to repress transcription from a distance.
TGIF Contains Multiple Repression Domains-To localize the repression domain within TGIF, a series of GBD/TGIF fusions were coexpressed with the TK-luciferase reporter containing Gal4p binding sites. As shown in Fig. 2A, both the amino-and carboxyl-terminal halves of TGIF (amino acids 1-137 and 138 -272) repressed transcription from the TK pro-  ). B, as described for A, except that in (Gal) 4 -TK-luc, transcription of the luciferase gene is driven by the herpes simplex virus thymidine kinase promoter. C, as described for A, except that luciferase gene transcription is activated by the adenovirus major late promoter. D, GBD/TGIF(1-272) was targeted to five Gal4p binding sites located either adjacent to the SV40 promoter or separated from it by 450 base pairs or 2.9 kilobase pairs. Luciferase activity was assayed as described in A and is presented as the mean Ϯ S.D. of triplicate transfections in which the activity of each reporter in the presence of a control expression vector has been set to 100%.
moter. Further deletion within the amino-terminal half of TGIF revealed that much of the homeodomain and sequences carboxyl-terminal to it appeared to be dispensable for transcriptional repression ( Fig. 2A). In contrast, deletion from the amino terminus of TGIF resulted in a clear reduction in the ability of this half of the protein to repress transcription ( Fig.  2A). These results suggest the presence of a transcriptional repression domain at the extreme amino terminus of TGIF.
A series of carboxyl-terminal fragments of TGIF fused to the GBD were coexpressed with (Gal) 4 -TK-luc. GBD/TGIF(138 -272) repressed transcription, and further division of this region revealed that it could be separated into two subfragments capable of repressing transcription: GBD/TGIF(138 -192) and GBD/TGIF(208 -272), each of which repressed transcription (Fig. 2B). Deletion of the carboxyl-terminal 42 amino acids from GBD/TGIF(108 -272) resulted in a slight reduction in repression. Creation of a small internal deletion within the GBD/TGIF(108 -230) construct resulted in a complete loss of its ability to repress transcription. These results suggest that the carboxyl-terminal half of TGIF constitutes a second repression domain, which may consist of two subdomains (Fig. 2C).
TGIF Represses Transcription when Bound Directly to DNA-Five copies of the TGIF binding site (CTGTCAA; Ref. 8) were inserted upstream of the TK promoter within TK-luc (creating (TG) 5 -TK-luc). As shown in Fig. 3A, expression of increasing amounts of TGIF in L17 cells reduced the activity of (TG) 5 -TK-luc by up to 80%. Little effect of TGIF expression on the TK-luc reporter lacking TGIF sites was observed. Thus, TGIF is able to repress transcription when bound directly to DNA via its homeodomain.
Coexpression of a series of TGIF deletion constructs with (TG) 5 -TK-luc revealed that the presence of the homeodomain is critical for repression by TGIF via its own site (Fig. 3B). However, expression of the homeodomain alone (construct 31-137) did not result in repression of transcription from (TG) 5 -TK-luc. All other TGIF deletion mutants containing an intact homeodomain repressed transcription to some degree, although some differences in transcriptional repression were observed (Fig. 3B).
To more carefully compare the contribution of each domain to transcriptional repression by TGIF when bound to DNA via its homeodomain, we tested five TGIF expression constructs that were expressed at similar levels (Fig. 3C). Increasing amounts of each expression vector were cotransfected with a TK-luciferase reporter containing two TGIF sites. In this case, repression by TGIF was less potent, due to the reduced number of TGIF binding sites. With the maximum amount of TGIF(1-272), activity was repressed to about 35% of that with a control vector (Fig. 3C). Carboxyl-terminal deletion to amino acid 192 had no effect on repression by TGIF. However, removal of an additional 28 amino acids (construct 1-164), leaving only the amino-terminal repression domain fully intact, dramatically reduced repression (Fig. 3C). A small internal deletion (construct 1-147:177-272) decreased transcriptional repression, suggesting that this deletion was targeting an important domain of TGIF. However, TGIF  was also unable to repress transcription as efficiently as the wild type protein, suggesting that the other domains of TGIF contribute to repression (Fig. 3C).

Repression by the Carboxyl-Terminal Half of TGIF Is
Dependent on Histone Deacetylase Activity-Repression of transcription by many proteins has been shown to involve HDAC activity. Repression of transcription from the TK promoter by GBD/TGIF(1-272) was examined in the presence of increasing concentrations of the inhibitor of deacetylase activity, TSA. As shown in Fig. 4, GBD/TGIF(1-272) repressed transcription to less than 5% of that seen with a control plasmid, and TSA had little effect. Because this fusion protein contains multiple domains capable of repressing transcription, the effect of TSA on one domain may be masked by the presence of a second repression domain. We therefore tested the effect of TSA on repression by GBD/TGIF(1-137) and GBD/TGIF(138 -272). TSA was unable to relieve repression by the isolated amino-terminal repression domain. However, transcriptional repression mediated by the carboxyl-terminal half of TGIF was significantly reduced by increasing concentrations of TSA (Fig. 4). These results suggest that the carboxyl-terminal region of TGIF may repress transcription by recruitment of a histone deacetylase, whereas the amino-terminal repression domain appears to be independent of this activity.
TGIF Interacts with Histone Deacetylases-To determine whether TGIF can interact with histone deacetylases, COS-1 cells were cotransfected with full-length TGIF tagged at its amino terminus with two copies of a HA epitope, together with Flag-tagged HDAC1. As shown in Fig. 5A, TGIF was present in Flag immunoprecipitates from cells expressing both HA-TGIF and Flag-HDAC1. In contrast, no TGIF precipitated in the absence of coexpressed HDAC. To identify the HDAC-interacting region within TGIF, a series of HA-tagged TGIF deletion constructs was coexpressed with Flag-HDAC1. Carboxyl-terminal deletion to amino acid 230 or amino acid 192 did not affect the interaction with HDAC. Similarly, removal of the aminoterminal 107 amino acids had no effect on the interaction (Fig.  5A). Deletion within the central repression domain of TGIF  1-230 and 1-147:177-230, Fig. 5A). Additionally, no interaction of amino acids 1-164 of TGIF with HDAC1 was observed. These results place the HDAC-interacting domain of TGIF between amino acids 108 and 192, suggesting that the central repression domain of TGIF acts by recruiting HDACs.
As shown in Fig. 5B, TGIF was clearly detectable in Flag immunoprecipitates from cells expressing Flag-HDAC1. Similarly, HA-TGIF coprecipitated with the related proteins (35) HDAC2 and HDAC3, suggesting that TGIF was able to interact with all three HDACs tested. DISCUSSION TGIF is a member of the atypical TALE family of homeodomain proteins (8,9). Members of this family of proteins have been shown to play roles in both transcriptional activation and repression. In the case of proteins such as Prep1, effects on transcription appear to be context-dependent. Thus, Prep1 is able to participate in either transcriptional repression or activation, depending on the other DNA-binding proteins with which it associates. Prep1 can displace HoxB1 and bind to DNA together with Pbx1, repressing transcription relative to a Pbx1/ HoxB1 complex. However, recruitment of Prep1 by a DNA -FIG. 3. TGIF represses transcription from its cognate site. A, increasing amounts of a TGIF expression construct were cotransfected into L17 cells with luciferase reporters in which transcription is activated by the herpes simplex virus TK promoter (TK-luc) or (TG) 5 -TK-luc, in which five TGIF-binding sites (CTGTCAA) are present upstream of the TK promoter. Luciferase was assayed as described in Fig. 1 and is presented as the activity relative to that obtained in the presence of a control expression vector. B, the activity of (TG) 5 -TK-luc was assayed as described in A in the presence of 200 ng of the indicated TGIF expression constructs. Activity is presented as described in A, Ϯ S.D. of triplicate transfections. C, L17 cells were transfected with a TK-luciferase reporter containing two TGIF binding sites and the indicated amounts of each TGIF expression vector. Luciferase activity is presented as the average of triplicate transfections. For each point, the S.D. was less than 10% of the mean and has been omitted for ease of presentation. The expression level of each of the five TGIF constructs from a parallel transfection into L17 cells is shown. Extracts were subjected to SDS-polyacrylamide gel electrophoresis and immunoblotted using a HA-specific antiserum. The positions of molecular mass markers (in kDa) are shown.

FIG. 4. Effect of trichostatin A on GBD/TGIF fusions. L17 cells
were cotransfected with the indicated GBD/TGIF fusions and the (Gal) 4 -TK-luc reporter. 18 h after transfection, TSA was added to the indicated concentration, and luciferase activity was assayed 18 h later. Activity is presented relative to that obtained from transfection of a control expression vector. bound complex of Pbx1 and HoxB1, without a requirement for DNA binding by Prep1, results in increased transcriptional activity (37,38). In contrast, TGIF appears to be an obligate transcriptional repressor. We have been unable to show transcriptional activation by TGIF, regardless of whether it was bound directly to DNA, tethered via a heterologous DNA-binding domain, or recruited to DNA by an activated Smad complex (28). However, we cannot rule out the possibility that interaction of TGIF with a specific protein would mask its ability to repress transcription. Under such circumstances, it is conceivable that TGIF could aid another protein in transcriptional activation, perhaps by stabilizing its binding to DNA.
TGIF appears to contain at least two functionally separable repression domains. The amino-terminal domain, including part of the homeodomain, represses transcription independent of HDAC activity. Repression by the amino-terminal region of TGIF is unlikely to be due to competition with activator proteins for binding to specific response elements because it is able to repress as a Gal4p fusion. Together with the fact that TGIF repressed several unrelated promoters, this also suggests that TGIF is not acting specifically on one transcriptional activator, but rather that repression by TGIF is at the level of more general factors.
Several homeodomain proteins repress transcription via interactions with components of the general transcription machinery. The Drosophila Eve protein represses transcription by interacting directly with the TATA-binding protein, preventing it from binding to the TATA element (39). It has been demonstrated that part of the homeodomain of human Msx-1 (termed the amino-terminal arm), which precedes helix one, interacts with TATA-binding protein and represses transcription (7). Similarly, repression by HoxA7 is mediated by the homeodomain and is modulated by the amino-terminal arm (6). The amino-terminal repression domain of TGIF may also repress transcription by interacting with TATA-binding protein or other general transcription factors. Alternatively, TGIF may interact with and recruit a second repressor protein, which confers the ability of this domain to repress transcription.
Within the carboxyl-terminal half of TGIF, two regions appear to contribute to transcriptional repression (see Fig. 6). The central region of the protein interacts with HDAC, whereas we have been unable to demonstrate an interaction of HDAC with the extreme carboxyl terminus of the protein alone. Experiments with an inhibitor of HDAC activity (TSA) demonstrated that, in contrast to the amino-terminal repression domain, repression by the entire carboxyl-terminal half of the protein is dependent on HDAC activity. One interpretation of this result is that both repression domains within this half of the protein contribute to the recruitment of HDAC. In isolation, the carboxyl-terminal region of TGIF can repress transcription in the context of a GBD fusion, but it does not stably interact with HDAC to allow detection by co-immunoprecipitation. If this region represents an entirely separate repression domain, it is likely that it would mask the effect of TSA on the central region, as the amino-terminal repression domain appears to do. However, we cannot rule out the possibility that TGIF contains three repression domains, only one of which is dependent on HDAC.
When bound to DNA via its homeodomain, full transcriptional repression by TGIF appears to require the central HDAC-interacting repression domain. Thus, a major way in which TGIF represses transcription may be by the recruitment of HDAC activity. However, the relative importance of these domains may be skewed by deletion analyses.
We have demonstrated that TGIF actively represses transcription from multiple TGIF binding sites. A tandem array of five TGIF-like binding sites is present within the retinoid X receptor response element from the cellular retinol-binding protein II promoter (8). Binding of bacterially expressed TGIF to this response element has been demonstrated, and this binding has been suggested to repress retinoid-dependent transcription. Our results demonstrate that such an element would be the target of active repression by TGIF. However, four of these TGIF-like sites are degenerate, suggesting that repression of this element by TGIF may require other factors. Interactions with other DNA-bound factors could modify the DNA binding specificity of TGIF. In light of numerous other homeodomain proteins that bind DNA cooperatively with other transcription factors (40,41), such a mechanism seems likely. Interestingly, many of these interactions involve multiple DNAbinding homeodomain proteins (5,37,38,42).
TGIF is able to repress Smad2/Smad3-dependent, TGF-␤activated transcription, and it appears that both the aminoand carboxyl-terminal repression domains are required for this corepressor activity (28). TGIF may play two separate roles within the cell: (a) acting as a specific corepressor for TGF-␤activated Smads, thereby limiting the effect of TGF-␤ signaling; and (b) repressing the transcription of a distinct set of genes. Thus, repression of transcription by TGIF, when bound directly to DNA, may have no functional overlap with its role in TGF-␤-activated transcription.
As shown in Fig. 6, TGIF appears to contain two separate recruitment domains: one that binds directly to DNA (8) and another, which binds TGF-␤-activated Smad proteins (28). Three other functional domains are present within TGIF: an amino-terminal, HDAC-independent repression domain (repression domain 1 (RD-1)), and two regions (RD-2a and RD-2b) within the carboxyl terminus of the protein that repress transcription in a HDAC-dependent manner. Although RD-2a alone is capable of interacting with HDAC, RD-2b may be part of the same carboxyl-terminal repression domain and may contribute to the formation of a TGIF/HDAC complex.
In summary, we have demonstrated that TGIF is an active transcriptional repressor. This repression is dependent, at least in part, on HDAC activity, suggesting that TGIF may regulate transcription by remodeling chromatin. TGIF represses transcription in a context-independent manner, suggesting that any promoter to which TGIF is recruited will be repressed. It will therefore be of great interest to identify the regulatory mechanisms controlling the level or activity of TGIF within the cell.